Sviluppo di metodi senza contatto per il monitoraggio di parametri fisiologici e attività nell'ambito dell'Ambient Assisted Living

One of the most interesting aspects in recent years thanks to which many projects have been developed, is Ambient Assisted Living (AAL). The aim of this thesis relies on this area, through the development of solutions for the monitoring of Activities of Daily Fiving (ADL) and physiological parameters using non-contact technologies. The subjects to whom the primary attention of the presented studies are addressed are elderly or frail individuals who have a constant need for observation and monitoring in their home environment, trying to preserve their autonomy. Independence for people suffering from motor or neuronal disorders is a determining factor and allows, from a socio-economic point of view, a saving of resources whose cost would weigh on the health system of the entire nation. In recent years there has been a great effort by the scientific community to be able to solve these problems through dedicated technologies, directly installed inside what are defined as ``smart homes". This work focuses on the development of algorithms and solutions for intelligent environments using RGB and depth sensors. Above all, we try to provide solutions that are not intrusive, comfortable and at the same time able to guarantee privacy. Methods aimed at monitoring and automating the control of correct nutrition based on neural networks, techniques for detecting and classifying falls for which depth images are processed, heart rate estimation from RGB images through face and skin recognition algorithms will be described, in which the main employed technique are the amplification of the skin color and the principal components analyses passing through different color spaces. Finally, with the aim of monitoring problems related to the central nervous system, an approach to extract the variation of the pupil diameter over time based on the development of segmentation and edge recognition techniques will be presented.